ApolloChoiceModelling forum

Dear all,

I am new to the forum and after finding the d-efficient design for my unlabelled experiment (two alternatives: 1 & 2), and collecting pilot data, I am now preparing for the choice modelling.

The five attributes & corresponding levels are the following:

Years of life (yol): 0.5, 1, 3
Quality of life gains (qol): 20, 40, 60
Experience (exp): Poor(0), Fair(1), Good(2)
Size (size): 50, 5000, 10000
Equity (equ): 25, 50, 75

Apart from interactions with socio-demographics, I want to add interactions between two of the continuous attributes yol & qol. I am trying the following in R, but it is not working:

### Vector of parameters, including any that are kept fixed in estimation
apollo_beta=c(asc = 0,
byol = 0,
bqol = 0,
bexp_fair = 0,
bexp_good = 0,
bsize = 0,
bequ = 0,
bage = 0,
byol_female = 0
)

apollo_fixed = c()

# ################################################################# #
#### GROUP AND VALIDATE INPUTS ####
# ################################################################# #

apollo_inputs = apollo_validateInputs()

# ################################################################# #
#### DEFINE MODEL AND LIKELIHOOD FUNCTION ####
# ################################################################# #

apollo_probabilities=function(apollo_beta, apollo_inputs, functionality="estimate"){

### Function initialisation: do not change the following three commands
### Attach inputs and detach after function exit
apollo_attach(apollo_beta, apollo_inputs)
on.exit(apollo_detach(apollo_beta, apollo_inputs))

### Create list of probabilities P
P = list()

### Create coefficients with interactions
b_years = byol + byol_female * (female==1)
b_yolqol1 = bqol + byol*yol1
b_yolqol2 = bqol + byol*yol2

### List of utilities: these must use the same names as in mnl_settings, order is irrelevant
V = list()
V[["Alt1"]] = asc + b_years*yol1 + b_yolqol1*qol1 + bexp_fair*(exp1==1) + bexp_good*(exp1==2) + bsize*size1 + bequ*equ1 + bage*age
V[["Alt2"]] = b_years*yol2 + b_yolqol2*qol2 + bexp_fair*(exp2==1) + bexp_good*(exp2==2) + bsize*size2 + bequ*equ2


OR should I simply add directly to the interactions between attributes to the utility functions (this one is working, but I want to confirm whether this is okay):
### Vector of parameters, including any that are kept fixed in estimation
apollo_beta=c(asc = 0,
byol = 0,
bqol = 0,
bqy = 0,
bexp_fair = 0,
bexp_good = 0,
bsize = 0,
bequ = 0,
bage = 0,
byol_female = 0
)

### Create coefficients with interactions (only with socio-demographics):
b_years = byol + byol_female * (female==1)

### List of utilities: these must use the same names as in mnl_settings, order is irrelevant
V = list()
V[["Alt1"]] = asc + b_years*yol1 + bqol*qol1 + bqy*((qol1)*yol1) + bexp_fair*(exp1==1) + bexp_good*(exp1==2) + bsize*size1 + bequ*equ1 + bage*age
V[["Alt2"]] = b_ years*yol2 + bqol*qol2 + bqy*((qol2)*yol2) + bexp_fair*(exp2==1) + bexp_good*(exp2==2) + bsize*size2 + bequ*equ2

I would appreciate very much your help with this. Thanks!

Best wishes,

Pamela.

Pamela

can you explain exactly what is happening when you say "it is not working"?

Stephane

Hi Stephane,

Thanks for your response.

I run it again and the model is actually working well.

The results are virtually the same, except for the value of the coefficient “bage” (although this is a synthetic variable).
When including interactions between attributes, should I use the first approach (i.e. defining the interactions before defining the utility functions) or the second one (i.e. defining the interaction within the utility function directly), or both are the same?

Many thanks!

Best wishes,

Pamela.

The two approaches should give the same results, so use whichever you find easier

Thanks, Stephane!

Best wishes,

Pamela.